Analysis: Predictive Analytics and Oracle OpenWorld

Predictive analytics seemed to sit it out at the recent Oracle OpenWorld, except for one recent Oracle acquisition.

B. G. "Buddy" Smith likes to tell about the system he built 15 years ago while in the Air Force. He also likes to say, as he scans the hundreds of booths near his own at Oracle OpenWorld, that he still doesn’t see much that can do what his system did. Back then, he called it a balanced scorecard. Now it would also be called predictive analytics.

"I don’t see process managed," he says with the trace of a drawl. Today he’s a senior consultant at ApexIT, a systems integrator and implementation consultancy based in Minneapolis. "If you fly an airplane by the gas gauge alone, you’re going to crash."

Smith was a full colonel in the Air Force when his boss, a four-star general, told him he needed to see at a glance what was going on at any moment within his command. The general oversaw 25 majors and about as many major installations in California, Alabama, Texas, and other locations around the country. His system monitored customer satisfaction, financials, processes, and people. Numbers flashed by on the big screen as new data came in. Like an onion, the overview peeled back to reveal a layer of detail, which itself peeled back for more detail, like today’s dashboards.

If satisfaction of internal clients, recorded first on paper, was moving downward at any base or unit, a call went out. The general wanted to know, "What’s causing it?" Maybe then he could head off something bad.

To learn how to design the system, Smith had gone away to Harvard for six months. Then he came back and built it on Cray mainframes. Now, 15 years later, he wonders why more organizations still don’t monitor leading indicators.

TDWI research confirms that most BI professionals have stayed away (see http://www.tdwi.org/Publications/WhatWorks/display.aspx?ID=8452). Predictive analytics helps choose good customers, sell them products and services, and then sell them even more. It can reduce churn. It’s a powerful budgeting and forecasting tool. Those are just a few known uses, and there are likely more to be found.

Sixty-six percent of those with such programs rated the business value to the company "high" or "very high." Yet only 21 percent of respondents in the August 2006 survey indicated that they had a program in place; 61 percent were just thinking about it.

In "Crossing the Analytics Chasm" in the March 2006 issue of TWDI’s Business Intelligence Journal (see http://www.tdwi.org/Publications/BIJournal/display.aspx?ID=7892), Rahul Asthana examined seven reasons companies use for holding off. For now, it can be boiled down to three: it’s scary, "we don’t need it," and "it won’t work here."

Smith and several others around the show floor added another reason: low awareness. "How do you ask for it if you don’t know it’s there?" said Smith. Cetova product manager Craig Foote said, "It’s hard to describe it when they haven’t really seen it."

Such a promising technology must have someone promoting it. I went looking.

After about a dozen conversations with talkative booth representatives, I spotted the Crystal Ball demo kiosk in the Oracle Demo Grounds. There, marketing communications specialist Kevin Weiner had an answer for Buddy Smith. "The market is very ready," he said, "Demand is definitely rising."

According to Weiner, "What we’re trying to do is take people from using a single-point estimate (which is deterministic) and to think in a range, to think stochastically." That is, don’t put all your chips on one number, spread them over a promising range. "What we’re doing is not very complicated," he said, noting that his software is intuitive and designed to nudge users off the single-point estimates and forecasts, numbers that are taken as reality in simple models.

The technology has many applications. A Mexican retail chain, whose representatives he met at the show, scouts 15 new stores every year. The chain now has 1500 locations around Latin America. For every new spot that they pick, they may dismiss five or ten. With Weiner’s software, they can compare a wide range of data, such as the characteristics of customers at their highest producing shops against those of prospective neighborhoods. The time saved could be enormous.

It may not be so easy to do. Some organizations are simply set in their ways, and some just don’t have the time.

iDashboards’s David Ferguson, who was promoting the company’s new pre-built views, cautions about the prep involved in predictive analytics and so many other solutions. One person who dropped by his booth said about software from a company that was not exhibiting at OpenWorld, "I had no idea that to get this pretty front end I’d have to do all this work on my data."

About the Author

Ted Cuzzillo, CBIP, is a freelance writer based in the San Francisco area. He can be reached at ted5@datadoodle.com.